Aren’t LLMs distinct from image generation and manipulation models?